comparative saving behavior of rural and urban … · the intertemporal nature of the saving...

Journal of Philippine Development I_tNumber Thirty One, Volume XVII, No. 2, 1990

COMPARATIVE SAVING BEHAVIOR OF RURALAND URBAN HOUSEHOLDS IN THE PHILIPPINES*

Romeo M. Bautista and Mario B. Lamberte**

Marginal saving rates for Philippine households are estima-ted, distinguishing between rural and urban households, byregion, and by income group. At a given income level, ruralhouseholds generally save more than urban households,both on average and at the margin. This contrasts with thehigher average saving rates for urban households in thevarious regions, attributable to their higher incomes. Theestimated marginal saving rates for rural households inmany regions are found to be higher than those for theirurban counterparts. In the context of agriculture-baseddevelopment, faster growth of rural incomes need notresult in lower aggregate savings.

I. INTRODUCTION

The aggregatesaving rate, the fraction of national income thatis not spent on current consumption, has long been widely regardedas a key factor in economic growth. 1 In the dynamic Harrod-Domarmodel, the saving rate and the incremental capital-output ratio joint-ly determine the growth rate of the economy. The critical role ofsaving in capital accumulation and economic development is alsorecognized in the "two_ap" and classicalgrowth models. Even in the

*Paper presented during the ACPC-PIDS sponsored seminar-workshop on "RuralSavings and Informal Credit in the Context of Financial Liberalization" held on June 7_,1990, at the G, Licaros Auditorium, 18th Floor, Multi-Storey Building, Central Bank of thePhilippines, A. Mabini Street, Manila.

*"The authors are Research Fellow at the International Food Policy Research Insti-tute (IFPRI) in Washington, D.C., and Vice-President of the Philippine Institute forDevelopment Studies (PIDS), respectively. The authors benefited from comments on anearlier draft of this paper by Dale Adams, Michael Lipton, and John Mellor, as well as fromparticipants in seminars at IFPRI, PIDS and Ohio State University (OSU). The fundingsupport from the United States Agency for International Development (USAID) throughthe NEDA-Training and Development Issues (TDI) Project, and the research assistance of

, Marion Alicando, are also gratefully acknowledged.

1, Past surveys discussingthe role of saving in economic development and its deter-minants include Snyder (1974) and Gersovitz (1988).

150 JOURNAL OF PHILIPPINE DEVELOPMENT

neoclassicalgrowth model in which savingsdo not influence econom-ic growth in the steady state, a higher savingrate is associatedwithmore rapid growth of the economy in its movement towards long-runequilibrium. As noted by Gersovitz (1988), this transitional path(which can take time) is more meaningful than alternative steadystatesin representingthe evolution of developingeconomies.

There are of courseother determinants of economic growth.Technological progress,institutional development, domestic policies,and the external economic environment have also been emphasizedin the development literature. To theseadditional influenceson eco-nomic growth can be attributed the lack of a simple correlation

• sometimesobservedbetween savingsand growth in developingcoun-tries (LDCs). Only when these explanatory variables are simulta-neously taken into account would one be able to assessempiricallytheir separate effects on growth. Although Arthur Lewis's famousdictum that raisingthe savingrate isthe "central problem in econom-ic development" can be disputed (Deaton 1989: 39), few woulddoubt that economic growth cannot be long sustainedunder condi-tions of decliningsavingrates.

Savings, including both domestic and foreign, finance the(physical and human) capital formation needed to increase output,

and this is of particular importance to typically capital-scarce LDCs.Apart from itsdirect contribution to output growth, capital accumu-lation also makes possiblethe employment of complementary pro-duction inputs in abundant supply -- for example, unskilled labor inmost developingcountries -- and servesas a vehiclefor the adoptionof improved technologiesembodied in new investments.

While there have been brief periods of significant inflow ofexternal financial resourcesto some LDCs in the past, foreign savingscannot be expected to provide a sustainable basis for financingdomestic investment. Raisingthe national savingrate is particularlyessential to developing countries with a heavy debt-serviceburdenand limited capacity to obtain loans in foreign capital markets.indeed, macroeconomic adjustment programs oriented to theresumption of long-run growth invariably emphasize the need toexpand domesticsavings.

Householdsaving is usually the largestcomponent of domesticsavingsin developing countries, especially the lower-income, predo-minantly agricultural LDCs. This contrasts with the much greaterimportance of corporate saving in developed countries. The ability,willingness, and opportunity of householdsto save over time cantherefore significantly influence the rate and sustainabilityof capitalaccumulation and economic growth in developingcountries.

BAUTISTA & LAMBERTE: SAVING BEHAVIOR 151

There has recently been a growing recognition of the pivotalrole of agricultureand the rural sectorin promoting a more rapid andequitable economic growth for low-income LDCs.2 Advocates of anagriculture-baseddevelopment strategy (ADS) emphasizethe expan-sion of public investments in the rural sector and the removal ofpolicy-induced price biasesagainst agricultural products. These areexpected to raise farm productivity, agricultural output, and ruralincome, which in turn will generate increasedintermediate demandfor nonagricultural inputs, and more importantly will stimulate con-sumption demand for food and labor-intensiveindustrial goodsandservices,generatingemployment and income multiplier effects on therural, regionaland national economies.In the short to medium term,rural household income is expected to increaseat a faster rate com-paredto urban householdincome.

How will the adoption of ADS and the associatedshift in rural-urban income distribution affect aggregatehousehold savings?Inany givenyear, it is commonly observedthat the averagesavingrateof rural householdsis lower, sometimesvery much lower, than thatof urban households.3 If fixed savingratesfor the rural and urbansectors are assumed,the rising real income of rural householdsrela-tive to urban householdsin a growth processthat is agriculturallydriven will likely .resultin a lower aggregatesavingrate, calling intoquestion its sustainability. Such an assumption on householdsavingbehavior, however, is questionable.

A positive relationship between the saving rate and incomein developing countries, at least within certain ranges of incomelevels, has been obtained in past empirical studiesusinghouseholdsurvey data (e.g., Bhalla [1980] for India) or cross-countrynationalincomeaccounts (e.g., Moore 1981 ) for Asian countries.An observedlower average saving rate for rural households (relative to urbanhouseholds) may then be explained simply by their lower averageincome. However, rural household incomes can increase rapidly inthe course of agriculture-baseddevelopment, which may prevent adecline in the aggregatesavingrate or evenraiseit. Furthermore, theimprovement of investment opportunities in the rural areasassocia-

2. See, for example, Adelman (1984) and Bautista (1988). An early statement ofan agriculture-based development strategy can be found in Mellor (1976),

3. Lipton (1977: 247) has aptly pointed out that, apart from the significant under-estimation of agricultural saving embodied directly in investment, "some rural saving isdrained off by price twists to finance socially Iow-yieldlng urban investment" and that"farmers would have more incentive to saveand to embody their savings in farm investment,if its returns were not artificially depressed by policies turning the terms of trade againstagricu!*.ure."


ted with the ADS providesan additional stimulus to increasedsavingsby rural households.

Empirical evidence on the relative size of the marginal savingrates for rural and urban householdsis thin and contradictory. Ina study using 1963-72 survey data on South Korean households,farmers were found to have been considerably more thrifty, withtheir marginalpropensity to consumebeingalmost half that of urbanconsumers (Lluch et al. 1977: 99). Based on a similar analyticalframework, estimates of the marginal propensity to save (MPS) forMexico show averagesof 0.11 for rural householdscomparedto 0.25for urban households;"in Chile the valuestend to be around 0.30,with slightly lower levelsfor comparable urban households" (p. 241 ).Usingdata from household surveysin Bangladeshfor each year from1976/77 to 1978-79, the MPS out of transitory income isestimatedto be "consistently and significantly higher" among rural households(Chowdhury 1987). On the other hand, the findings from an Indianstudy usingtime seriesdata indicate a higher aggregateMPS for theurban sector (Gupta 1970). To be sure, comparability of savingrateestimates from independent studiesfor different countries by diffe-rent investigators is severely impaired by differences in data andmeasuresused and by the varying analytical approachesand estima-tion techniquesadopted in derivingthe estimates.4

In this paper, we investigate the comparative savingbehaviorof rural and urban householdsin the PhilippinesusingFIES (FamilyIncome and Expenditure Survey) data for 1985, the latest year forwhich such data are available. Apart from the contribution to empi-rical knowledge in an area previously not given systematicattentionin the development literature, some policy interest attaches in thePhilippine context to an examination of the differential savingbeha-vior of rural and urban households considering the recent shifttowards more rural-oriented development policiesasexpressedin thegovernment's Medium-Term Philippine Development Plan, 1987-1992.

Section II addressessome theoretical and empirical issuesinsavingsbehavior, especially as they relate to existing conditions indeveloping countries and, in particular, the Philippines. The discus-sion leads to the specification of household savingfunctions distin-guished by region, by location (rural and urban), and by incomeclass. In Section III, we describe and evaluate the data basefor the

4. For instance, the use of an "extended" linear expenditure system and of sophis-ticated econometric estimation procedures in the studies on Korea, Mexico and Chile citedabove contrasts sharply with the simple linear regression of income on saving employed inthe studies on Bangladesh and India.

BAUTISTA & LAMBERTE; SAVING BEHAVIOR 153

study, indicating also some relevant characteristicsof sample house-holds. Section IV presentsand analyzes the estimated savingequa-tions. The findings of the study are summarized and their implica-tions for development policy and strategyare briefly discussedin theconcludingsection.

I1. ANALYTICAL CONSIDERATIONS

In a fundamental sense,one'ssavingisa sacrificeof current con-sumption that will allow for an increase in future consumption.Saving is a means not only of reducingfluctuations in income andsmoothing consumptionover time but alsoof earning interest. Giventhe intertemporal nature of the saving process, it is the lifetimeincome of the individual, and not just the currect incomeas impliedin the Keynesian consumption function, that should influencecurrent saving. This life-cycle model also suggestsa higher rate ofsavingduring certain periods,e.g., in preretirement years in order toprovide for consumption in old age. The ageof the individualwouldtherefore be an appropriate determinant of savings.Additionally,existing investment opportunities (or, more precisely, the return toinvestments)and the cost of borrowing can also significantly affectsavingbehavior.

That current saving depends on lifetime resourcesis also animportant implication of Friedman's (1957) permanent incomehypothesis. Current income is viewed as the sum of permanent andtransitory incomes, the former reflecting the individual's lifetimeearnings. In itsextreme form, the permanent income approach postu-latesequality between an individual's current consumptionand per-manent income, implying that the MPS is zero out of permanentincome and one out of transitory income. Existing empirical studiesdo not bear out this strict versionof the permanent income hypothe-sis,but they providesupport to the view that the marginalsavingrateis higherout of transitory income than out of permanent income.

Quantitative information on actual choicesabout savingsistypically available at the household rather than individual level, atleast among developing countries. Such data are often providedthrough countrywide income and expenditure surveys, and lessfre-quently through special, Iocationally more focused surveys onchanges in family assets and liabilities. Both sources of data aresubject to error. Household income, particularly the nonmonetizedcomponent, is often observedto be underreported, with significantdifferences in the degree of underestimation among surveysdone inthe same country for different years (Berry 1985). If nonmonetized

154 ' JOURNAL OF PHILIPPINE DEVELOPMENT

consumption is also understated, the saving rate (but not necessarilythe level of savings) will be overestimated. On the other hand,household assets and liabilities tend to be incompletely enumerated(data on cash and jewelry are especially difficult to obtain) and non-financial investment is frequently not properly valued.

Given the limitations of the data used, some analysts havesimply examined the determinants of household savings on a variable-by-variable basis (e.g., Alamgir 1976), instead of using a systematicapproach to the modeling of saving behavior. While income is gene_rally acknowledged as the principal influence on saving, there areconceptual and/or data-related difficulties, in a developing countrycontext, in representing the income variable by current income or bysome measure of lifetime savings. In this study, estimation of house-hold saving functions makes use of the two alternative measures ofthe income variable.

The above representation of intertemporal decisionmakingabout individual saving needs modification if the household is to bethe unit of analysis as required by data availability. One complicatingfactor is the possible influence of family size on savings. Other thingsthe same, the higher the proportion of household members thatconsume more than they produce, the lower will be the householdsaving (Left 1969). Family size as such would not be the relevantexplanatory variable; among rural families particularly, even childrencan contribute significantly to the household's production andincome. Some measure of "dependency" reflecting the unemploy-ment of household members would be more appropriate.

The strong retirement motive for saving in the life-cycle hypo-thesis can also be called into question in the context of the strongfamily ties that characterize many LDC households. Especially in therural areas where the extended familysystem is more prevalent, thereis a sense of obligation to care for the older, economically inactivehousehold members. Thisalso reduces the need for the youngermembers to save since their future consumption is expected to beprovided for (at least partly) and weakens the expected relationshipbetween age (of household head) and savings.

Many households in developing countries are observed to havea low saving rate, or even a negative one, at low income levels. This issometimes interpreted to indicate problems of survey data reliability.Alternatively, however, the low or negative saving rate may reflectrational household responses to current, transitorily low incomes, orto the high consumption needs of the poor since "current consump-tion is more likely to influence survival and efficiency-at-work at lowlevels of consumption" (Gersovitz 1988: 410).


As pointed out above, there is some.empirical evidence of risingsaving rates as income level increases. As a prominent example,Bhalla (1980) has estimated a nonlinear savingfunction for ruralhouseholdsin India in which the averagepropensity to save is zero atthe subsistencelevel and increasesat an acceleratingrate in the low-income range, followed by a deceleration and eventual tapering offto an asymptotic value. However, asin the formulation of other non-linear saving functions (e.g., quadratic or semilogarithmic), theprocessor mechanismthat leads to the nonlinear relation betweenthe savingrate and income level is not spelled out. Due to deficien-cies in the data used, the view has been expressed recently that"hypotheses about behavioral nonlinearities in savings" cannot bedisentangledfrom "problems in measuringthe variables" (Gersovitz1988:411 ). Alternatively, one can estimate a linear savingfunctionfor (homogeneous)householdsdifferentiated by income group andthen compare their estimated saving propensities. This is the ap-proachadopted in the presentstudy.

In many developing countrieswhere capital marketsare signifi-cantly fragmented (McKinnon 1973), households face differentinvestment opportunities and costsof borrowing that could lead todifferences in marginal savingrates. Lower-income householdstendto be more vulnerable to capital market imperfections, and this isattributable in large measure to their weak information base andto their inability to meet collateral requirements. In the Philippines,the differing locations of householdsin geographicareasseparatedby wide spaces and having different consumption and productionpatterns are like!y to imply varying rates of return to investments.Access to credit likewise differs, influenced in part by regionalvariations in the effectivenessof financial intermediation. In addi-tion, insofar as "taste" is concerned (pure time preferences in con-sumption), some ethnic classesin certain regions are traditionallyknown either for their frugal or spendthrift ways. It would be appro-priate, therefore, to distinguish household savingfunctions by geo-graphicregions.

Within a given region it is necessaryfor purposesof this studyto distinguish rural and urban householdsin their savingbehavior.Apart from likely differences in demographic characteristics,social

practices, and educational background, rural households are ofcourse much more engagedin agriculture and face greater incomevariability relative to urban households. It issometimesassertedthatthe consumption pressuresof the demonstration effect are weakerin the rural areas, where the scope for conspicuousconsumption ismore limited so that rural families will save more compared with


their urban counterparts. Contrary to this, the findings of a studyfor Taiwan indicate that increased awareness and ownership ofmodern consumption goods led to higher savings,and that "thisrelationshipheld within incomegroups" (Freedman 1970:31 ).

It has been argued (although contrary currents of argumentalso exist) that there are more immediate outlets for investmentin the farm than in the cities and that if interest paid and the rate ofreturn in the rural sectorare higher than thoseavailableto the urbanpopulation, then farmers' marginal propensity to save will behigher. Also, basedon the permanent income hypothesis, the rela-tive instability of farm income would imply greater savingout ofcurrent income for rural households than for urban households.

Some of the above considerationsrelate not to the rural-urbandistinction which is Iocational, but to the differentiation of house-holds by source of income. Accordingly, it would be useful to dis-tinguish also between farm and nonfarm households, especiallysince the proportions of farm (nonfarm) householdsin urban (rural)areasare not insignificant (seebelow).

II1. THE DATA BASE AND PROFILE OF SAMPLEHOUSEHOLDS

As indicated above, the primary source of data for this study isthe Family Income and Expenditure Survey (FIES) for 1985. TheFIES is supposed to be conducted by the National Census andStatistics Office every five years beginning 1961. However, after thethird survey in 1971, it was only in 1985 that the next FIES wasundertaken. While previous surveyswere carried out through onlyone round of interviews, the 1985 FIES entailed two visits byenumerators (in July 1985 and January 1986), which obtainedinformation for the first and second halves of the year. Extensivereinterviews were also conducted subsequently to follow up onseeminglyquestionable survey responses.Additionally, some qualita-tive improvements were implemented in the 1985 FIES in dealingwith nonsampling errors and in the inclusion of noncash incomeand expenditures. It isalsoworth noting that the 1985 FIES and theearlier surveysdiffer in the numberof regionsinto which the samplehouseholdsare classified and in the regional grouping of provinces,which precludes direct comparability of regional data from thosesurveys.5

5. For an extended disc_,ssionof the reliability of 1985 FIES data and a comparisonwith other sources of savingsdata, see Oshima (1988) and Lamberte and Bautista (1989,Ch. IV).


The sample consistsof 17,495 households.They were selectedusing a stratified two-stage cluster sampling design, with villagesor "barangays" (classifiedas either urban or rural) as the primarysampling units and householdswithin each sample barangayas thesecondarysampling units. The samplingfraction was typically 1:400for urban areas and 1:600 for rural areas, with special samplingfractions applied to relatively small areas.A total of 16,971 samplehouseholdswere successfullyinterviewedin the two visits.

The distribution of sample householdsby regionand by urban-rural classificationis given in Table 1. (SeeAnnex I for the namesofregions and the provinces and chartered cities comprising eachregion;and Annex II for the classificationof rural and urban house-holds in the 1985 FIES.) Since this study also examines the savingbehavior of farm and nonfarm households,the samplehouseholdsineach region are also classifiedin the table accordingto whether theirincome is derived mainly from agricultural or nonagricultural activi-ties.

Rural households constitute some 53 percent of the totalsample households. It isworth noting that 42 percentof rural house-holds are classified as nonfarm households.While the proportionvariesby region (ranging from 25 to 54 percent), this indicatessub-stantial nonagriculturalactivities in the rural areas. As regardsurbanhouseholds,a not insignificant proportion (11 percent) is engagedin farm activities, also varying by region (from 7 to 32 percent).The lack of substantial correspondence between rural and farmhouseholdsand between urban and nonfarm householdssuggeststheusefulnessof estimating savingfunctions separately for these fourhouseholdcategories.

The mean values of household income, savings and savingrate by region, location and main source of income are showninTable 2. The large differences in averagehousehold income acrossregions -.especially those relating to the National Capital Region(NCR), the highest income region - reflect the past unevendevelop-ment of the Philippine economy and the geographicconcentrationof incomegrowth. Region III, which is located close to the NCR,ranks second.The averagehousehold income in Region VIII, whichJsthe most depressedin the country, is lessthan 30 percent of thatin the NCR. Within each region, a wide disparity between rural andurban average household incomes can be observed, with urbanhouseholds consistently showing higher values in all the twelveregions. A similar disparity is found if households are classifiedaccordingto their main sourceof income. In particular, the average


Table 1DISTRIBUTION OF SAMPLE HOUSEHOLDS:BY LOCATION AND BY MAIN OCCUPATION

Rural Urban

Region Sub- Sub- TotalFarm Nonfarm total Farm Nonfarm total

I 381 444 825 36 379 415 1240

(46.2) (53.8) (100.0) (8.7) (91.3) (100.0)

II 410 207 617 73 158 231 848

(66.4) (33,6) (100.0) (31.6) (68.4) (100,0)

III 366 474 840 62 809 871 1711(43.6) (56.4) (100.0) (7.1) (92.9) (100.0)

IV 593 618 1211 160 988 1148 2359(49.0) (51.0) (100.0) (13.9) (86,1) (100.0)

V 438 314 752 50 268 318 1070

(58.2) (41.8) (100,0) (15,7) (84,3) (100.0)

V l 654 307 961 89 465 554 1515(68.0) (32.0) (100,0) (16.1) (83.9) (100.0)

Vll 387 404 791 70 463 633 1324(48.9) (51.1) (100.0) (13.1) (86.9) (100.0)

Vlll 436 193 629 56 201 257 886(69.3) (30.7) (100.0) (21.8) (78.2) (100.0)

IX 411 182 593 36 156 192 785(69.3) (30.7) (100.0) (18.8) (81.2) (1OO.0)

X 374 236 610 74 254 328 938

(61.3) (38.7) (100.0) (22.6) (77.4) (100.0)

X I 481 164 645 107 383 490 1135(74.6) (25.4) (100.0) (21.8) (78.2) (100.0)

Xll 359 182 541 40 175 215 756

(66.4) (33.6) (100.0) (18.6) (81.4) (100.0)

NCR .... 2404 2404 2404

(100.0) (100.0)

TOTAL 5290 3725 9015 853 7103 7956 16971

(58.7) (41.7) (100.0) (10.7) (89.3) (100.0)

Note: Figures in parentheses are percentages of total rural or total urbanhouseholds. See Annex I for the names of regions and provinces/chartered cities in each region.

BAUTISTA & LAMBERTE: SAVING.BEHAVIOR 159

Table 2

MEAN VALUES OF HOUSEHOLD INCOME (Y), SAVINGS (S),AND SAVING RATE (SLY): BY REGION, LOCATION

AND MAIN SOURCE OF INCOME

Region/Households Y S S/Y(P) (P) {%)

I. All Households 32194 6732 20.9Rural 27490 5689 20.7.Urban 41544 8805 21.2Farm 26780 5722 20.4

No nfarm 40818 8720 21.4

i1. All Households 29286 6324 21.6Rural 24950 5513 22.1Urban 40867 8490 20.8Farm 18305 4160 22.7Nonfarm 31087 4459 14.4

III. All Households 40439 6664 16.5Rural 28750 3158 11.0Urban 51712 10045 19.4Farm 18162 1998 11.0Nonfarm 37346 , 6909 18.5

IV. All Households 31478 5255 16.7Rural 24073 3967 16.5Urban 39289 6614 16.8Farm 15764 1472 9.3Nonfarm 34903 6441 18.4

V. All HousehoIds 21506 2619 12.2Rural 17089 1510 8.8Urban 31953 5242 16.4Farm 13924 1496 10.7Nonfarm 27374 4037 14.8

V I. All Households 26699 4465 16.7Rural 19027 2537 13.3Urban 40007 7809 19.5Farm 15530 1899 12.2Nonfarm 31131 5521 17.2


Table2 (continued)

Region/Households Y $ SlY(P) (P) (%)

V I I, All Ho usehoIds 21758 4902 22,5Rural 15465 3076 19.9Urban 31096 76'12 24.4Farm 11261 2016 17.9Non farm 25488 6391 25.1

VIII. All Households 18666 2811 15.1Rural 15568 1842 9.9Urban 26249 5182 19.7Farm 11916 1247 10,5Nonfarm 26884 5845 21.7

IX, All Households 24094 5226 21.7Rural 22426 5128 22,9Urban 29254 5529 18.9Farm 15013 2274 15.2Nonfarm 29192 6516 22.3

X. All Households 27787 6265 22.6Rural 21729 3535 16.3Urban 39054 11342 29.0Farm 26598 6620 24.9Nonfarm 29107 3834 13.2

X I. All Households 29210 6004 20,6Rural 21831 3416 15.6Urban 38924 9409 24.2

Farm 23195 3376 14.6Nonfarm 32569 7519 23.1

X lI. All Households 25940 --3660 -14.1Rural 21216 2499 11.8

Urban 37829 -6583 -17.4Farm 17411 3442 19,8No nfarm 32434 -2818 -8.7

NCR All Households 64449 12790 19.8qLLl__

Source: Calculated from basic data in the 1985 Family Income andExpend itures Survey,


income of farm households is about one-half that of nonfarm house-holds in almost all the regions.

All regions, except one, show positive average householdsavings. As indicated in Table 4 below, the definition of savingsusedin this study includes expenditures on equipment, consumer durablesand education. The household saving rates vary considerably acrossregions, ranging from -14 to 23 percent. The Jrelatively high savingrates in Regions I and II are not surprising, considering that thepopulation there (mostly, the "llocanos") is traditionally knownfor thriftiness. The other regions that show relatively higher savingrates are Region VII, where the second premier city of the countryis located, and Regions IX, X and Xl, all from the Mindanao area,which are heavily populated by migrant families. There is a consi-derable difference in the saving rates between urban and rural house-holds within the same region. In particular, the former's savingrates are higher than those of the latter in nine regions.

The differences in saving rates between farm and nonfarmhouseholds within the same region are also quite substantial. Thesaving rates of nonfarm households in nine regions are higher thanfarm households. This situation occurs in almost all regions wherethe saving rates of urban households are found to be higher thanthose of rural households.

Table 3 compares the average saving rates between rural andurban households belonging to the same income bracket. Two impor-tant observations should be noted. First, urban households dissaveif their annual income falls below 1_20,000, while rural householdsdissave if their annual income is below 1_15,000. The differentialcut-off income for dissaving between rural and urban householdscould be due to the higher cost of living in the urban areas.Second,for the same income classthe saving rates of rural householdsarehigher than those of urban householdsin all but one income class.6This would seem to suggestthat the lower savingrates observedacross regions for rural households can be attributed at least inpart to their lower incomescomparedto urban households.

IV. ESTIMATION RESULTS

Our preliminary regressionsindicate the lack of significanceand, sometimes, theoretically incorrect signs of the estimated co-efficients of the following explanatory variables: age of household

6. Lipton (1977, Ch, 10) cites similar evidence of higher rural saving rates in Pakistanand India.


Table 3HOUSEHOLD SAVING RATES BY INCOME CLASS AND LOCATION

Incomeclass Philippines Urban Rural(in pesos) (%) (%) (%)

Under 2,000 --32 --78 --212,000 -- 3,999 --36 --86 --304,000 - 5,999 -22 -22 -226,000 - 7,999 -14 -17 -14

8,000 - 9,999 -8 - 12 -710,000 - 14,999 -2 -7 -115,000 - 19,999 3 -2 420,000 - 29,999 5 2 730,000 - 39,999 10 8 1240,000 - 59,999 13 11 1660,000 - 99,999 17 15 23

100,000 - 249,999 24 33 24250,000 - 499,999 32 33 24

500,000 and over 65 64. 80

Total 13 16 10

Total no of households 95,663 36,024 59,639

Source: Lamberte and Lira (1987).

Note: Saving is defined as total income minus total expenditures.

head, entered in quadratic form; educational attainment of house-

hold head, distinguishing among five education categories; and.

wife's employment status (employed or unemployed). In mostcases these variables are highly correlated with household income,

so that their separate effects on savings cannot be disentagled.

Accordingly, these variables .have been excluded in subsequent re-.

gressions.

Table 5 presents the estimation results for "all" households

in each of the 13 regions distinguished in the 1985 FIES, based

on the Keynesian saving function in which the income variable is

measured as current household income (after taxes). From 48 to

95 percent of the variance in regional household savings is explained

in the various estimated equations. Without exception the coeffi-

cient estimates for the income variable are highly significant (at

the 1 percent level). They range widely from 0.334 (Region II)

to 0.775 (Region X), indicating that an aggregate saving rate for

Philippine households would mask large differences in regional


Table 4DEFINITION OF VARIABLES

Notation Definition

Y Household disposable income in pesos( = household current income lesstaxes)

YP household permanent income in pesos

YT Household transitory income in pesos

5 Household savings in pesos ( = Y lesstotal)household expenditures net of expenses ondurable furnitures, equipment, and education)

LOC Location (1 for urbanhousehold; 0 for ruralhousehold)

D Main occupation of household head (1 fornanfarm household; 0 for farm household)

DR Dependency ratio ( = number of unemployedhousehold members divided by householdsize)

saving propensities. It is worth noting that the MPS estimate forthe National Capital Region (Metro Manila) is the second lowestamong the 13 regions in either of the two specifications.As expec-ted, the coefficient estimatesfor DR (dependencyratio) are negative,but in some casesthey are not significantly different from zero.7

Distinguishing between rural and urban households in eachregion, the estimated savingequations in Table 6, also basedon thecurrent income model, show some significant differences in MPSestimates for the two household classes.Rural households in Re-gions I and IX have markedly higher coefficient estimates for theincome variable compared to urban households,while the oppositeis true for Regions iV, VII, X, XI and XII. It is not possible,there-fore, to make a generalized inference on the relative savingpropen-sitiesof rural and urban householdsin the Philippines.

The MPS estimates for rural households, which range from0.322 (Region VII) to 0.735 (Region IX) in the specification without

7. As John Mellor hassuggestedin a privatecommunication,the composition,notjust the number, of unemployedhouseholdmembers(the numerator in the dependencyratio) would be a relevantfactor. Becauseeducationexpenditurecounts assaving,morechildren of school-goingage should imply highersaving--which'counterbalancesthe ex-pectednegativeeffectof DR.


Table 6

ESTIMATED SAVING EQUATIONS, CURRENT INCOME MODEL:ALL HOUSEHOLDS BY REGION

Region Const. Y /?2 : Const. Y DR R2

I -7353 0.438 0.70 -6389 0.438 -1418 0.70

(-18.6)* (53,4)* (-6.3)* (53.4)* (-1.0)

Ii -3460 0.334 0.48 -2074 0.335 -2233 0,48(-7.0)* (27.9)* (-2,0)** (28,0)* (-1.5)

III -13289 0.498 0,66 -12411 0,493 -1272 0.66(-27.7)* (58.0)* (-10.3)* (58.0)* (-0.8)

IV -9387 0,465 0 66 -8008 0.466 -2104 0,66

(-29.9)* (68.2)* (-10.7)* (68.2)* (-2.0)**

V -5966 0.399 0.54 -4194 0.400 -2733 0.54

(-16.8)* (35.5)* (-5.0)* (36.6)* (-2.3)*

VI -10419 0,568 0,77 -8562 0.557 -2934 0.77

(-24.6)*' (71.8)* (-7.8)* (71.7)* (-1,8)

VII -5019 0.456 0.72 --2525 0.458 --4100 0.73

(-17.5) (588)* (-3.9)* (59,3)* (--4.3)*

V III --5052 0.421 0.60 -4066 0.422 -1569 0.60

(-15.8)* (36.2)* (-5.7)* (36.3)* (--1.5)

IX -11600 0.698 0.86 --9697 0.699 -2744 0.87(--29.9)* (70.8)* (--8.5)* (71.0)* (--1.8)

X --15266 0.775 0.95 -13612 0,775 --2497 0.95

(--32.7)* (128.1)* (--9,3)* (128.1)* (--1.2)

X l -13664 0.673 0.86 -10788 0.673 -4244 0.86

(-30.7)* (82.4)* (--8.1)* (82.5)* (-2.3)**

X II -10954 0.563 0.75 -8944 0.563 --2853 0.75(-22.9)* (47,1)* (-5.9)* (47.2)* (-1.4)

NCR -12352 0,380 0.73 -7194 0.380 -7354 0.74(-162)* (82.1)* (-2.8)* (82.0)* (-2.1)**

Notes: See Table 4 for definitions of symbols, Numbers in parentheses aret-values,

R2 denotes adjusted coefficient of determination.

* Significant at the one percent level,** Significant at the five percent level.


the dummy variable for main source of income, may seem quitehigh. Nearly all these estimates, however, are even lower than theestimated MPS of 0.728 obtained in a recent study of rural savingsin the Philippines (Rodriguez and Meyer 1988) using a differentdata set (based on a special survey involving 980 rural householdsin six provinces).

Also from Table 6, a significant negative influence of thedependency ratio is seen for rural householdsin most regions;s inthe caseof urbanhouseholds,only one region (V I I) showsa statistic-ally significant coefficient estimate for DR. A possibleexplanation isthat, compared to urban households, there are greater opportunitiesfor farm work among the very young and very old members of ruralhouseholds, but lessopportunity for spending the additional income.That the marginal saving rates differ by source of income for ruraland urban households in several regions is indicated by the statistic-al significance of the coefficient estimates for the interaction termD*Y. They are mostly positive for rural households, implying thattheir MPS out of nonagricultural income is higher than that out ofagricultural income.9

Tables 7 and 8 contain the estimated saving equations using thepermanent (lifetime) income specification. As discussed in AnnexIII, the estimation of permanent income is based on the hypothe-tical earning capacity of households, determined from the estimatedrelationship for each region between household income and variousindicators of the stock of human, physical and financial assets.Tran-sitory income is derived residually, after subtracting permanent in-come from disposable income. The coefficient estimates for bothpermanent and transitory incomes are seen to be statistically signi-ficant, almost always at the one percent level. It is also remarkablethat the estimated MPS out of transitory income is higher than thatout of permanent income in all but one of the estimated equations.Higher values of the adjusted coefficient of determination are shownin Tables 7 and 8 compared to those in Tables 5 and 6, indicatingthat a larger proportion of the variance in savings is explained byconsidering the separate influences of the permanent and transitory

8. The coefficient estimate for the dependency ratio has also been found in theRodriguez-Meyer study to be negative; it is not statistically significant in the current incomespecification of the saving equation, but significant at the 5 percent level basedon the per-manent income model.

9. Similarly, Bhalla (1980) obtained generally lower MPS estimates for Indian ruralhouseholds out of agricultural income relative to other income sources. Based on time-series

national income accounts data, Burkner (1981) found a significant positive relationshipbetween household savingsand the ratio of industrial to agricultural income for the Philip-pines but a negative relationship for Thailand,

Table 6 _,ESTIMATED SAVING EQUATIONS, CURRENT INCOME MODEL: • o_

RURAL AND URBAN HOUSEHOLDS BY REGION

Reg ionlHouseho|d Const. Y DR R2 Const. Y D R D D* Y R 2

Region I :Rural --8352 0.561 -2024 8.76 -2707 8.317 --1864 --7486 0.273 0,78

(-7.9)* (51.2)* (--1,4) (-2.1)** (6,2)* (--1,3) (--7.0)* (6.8)*

Urban --4579 0.358 -2100 0.70 --640 8.204 --2168 --4112 0.155 0.78(-2.4) * (30.8) * (-0.8) (-0.1 } (0.9) (--0.6) (--1.8) (0.2)

Region I l:Rurat -2498 0.364 --1686 0.55 --4107 0.443 --1667 2582 --0.125 0.57

(-3.0)* (27.7)* (--1.4) (--4.5)* (20.9)* (--1.4) (2.9)* (-4.6)* C

Urban -2686 0.331 -3815 0.43 --4896 0.432 --3932 244.1 --0.103 0.43 ::0Z

(-0.8) (13.2)* (-0.8) (-1,0) (3.0)* (-0.9) (0.6) (-O.7) >rO

Region III: -nRural -7489 0.475 -4346 0.52 --4192 0.354 -3483 -6292 0.169 0.54 T

(-6.4)* t30.4}* (-2.8)* (3.2)* (10.6)* (-2,3)** (--5.3)* (4.5)* __F9"10

Urban --17782 0,518 1507 0.70 -9879 0.321 1586 --8264 0.201 0.70 -oZ(-8.5) * (44.5) * (--O.5} (-2 2)* (3.0)* (0.6) (-2.0)* * (! .8) rn

_2

Region IV: .: m• <Rural -282 t 0.362 -2944 0.49 -3369 0.434 -3119 198 -O.076 0.50 m

• I--

(-3.9)* (34.1)* (-3.0)* (-4.0)* (t6.3)* (-3.1)* (0.2) (-2.6)* o

Urban -13380 0.514 -323 0.73 -5889 0.313 -6 -8652 -O.210 6.73• rn

(-10.3)* (555)* (-0.2) (-2.5)* (3,6)* , (-0.0) (-4,0)* (--2A)** z. --p

Table6 (continued) m_>C-I

Regiot'_'HousehoId Con_. Y DR /72 Const. Y DR D D * Y R2

Region V:

Rural -3033 0.417 -3928 8.55 -286 0238 -3764 -4424 0.232 0.57 _,(-4.4)* (30.0)* (-4.1)* (-0.4) (7.5)* (--4.1)* (-6.6)* (6.0)* q00

Urban -6986 0.413 -1528 0.54 1205 0.300 -1784 -8767 0.399 0.55 :o

(-3.2)* (19.3)* (--0.5) (0.35) (03) (--0.6) (-2.9)* (3.5)* m

Region V I: :]><Rural -4462 0.429 -1908 0.64 -3732 0.404 -1864 -2062 0.048 0.64

(--6.5)* (412)* (-1.9)* (-5.2) (23.2)* (-1.9)* (-3.1)* (2.2)** c_gO

Urban -15732 0.597 -506 0.81 -6711 0.476 -84 11219 0.138 0.81 _:(--5.4)* (48.0)* (-0.1) (-1.9) (12.4)* (-02) (-4.3)* (3,4)* _"<

Region V II : "-nRural --1312 0.a22 -975 0.63 -823 0.238 -984 -58 0.086 0.63

(-2.9)* (36.5)* (-1.5) (-1.4) (7.6)* (-1.5) (-0.1) (2.6)*

Urban -3564 0.518 -7740 0.77 •-2242 0.313 0.7111 -6977 0.212 0.78(-2.4)* (42.5)* (-3.7)* (0.8) (1.9) (--3.4)* (-2.6)* (1.3)

Region VIII:RuraJ -3375 0.400 -1596 0,56 -2572 0.336 -1 487 -1504 0.085 0.56

(-5.6)* (28.4)* (-1.9) (-3.7)* (10.2)* (-1.8) (-2.2)* (2.3)**

Urban -5664 0.444 -1227 0.61 -2352 0.252 - 1002 -3863 O.199 0.61(-2.8)* (19.9)* (-0.4) (-0.3) (1.8) (-0.4)(-1.4) (1.4)

Region tX:Rural -8403 0.735 -4271 0.91 -2973 0.506 -4549 -10463 0.273 0.93

(-2.7)* (16.6)* (-3.3)* (-2.7)* (16.6)* (-3.3)* (-12.3)* (8.5)*

Table 6 (continued)

Region/Household Const. Y DR R2 Const. Y DR O D* Y R2

Urban -7150 0.450 -672 0.53 -3855 0.299 -440 -4127 0.168 0.33(-2.3)** (14.4)* (-0.2) (-1.0) (2.6)* (-0.1) (-1.4) (-1.4)

Region X:Rural -3635 0.415 -2752 0.65 -2819 0.393 -2803 -2413 0.860 0.65

(-4.4)* (33.4)* (-2.4)** (-3.3)* (22.9) (-3.2)* (-3.2)* (2.3)**

Urban -18685 8.796 -1615 -0.96 -105 0.I57 -1414 -21065 0.644 0.97

(-5.7)* (02.2)* (-0.3) (-0.0) (i .0) (-0.3) (5.9)* (4.2)

Region X I:Rural - 1459 0.323 -3220 0.50 -1770 0.369 -3582 103 -0.068 0.52

(-2.1)** (25.6)* (-3.4)* (-2.4)** (19.7)* (-3.8)* (0.2) (-2.6)*

Urban -16649 0.711 -2432 0.89 -7134 0.495 -2727 -10536 0.223 0.89 c(6.1)* (63.0)* (0.6) (-1.9) (6.0)* (-0.7) (-3.6)* (2.7)* z

Region X II : rRural -69 0.332 -6328 0.44 -549 0.443 -7269 -16 0.127 0.47 o"11

(-0.1) (19.9)* (-4.9)* (-0.5) (15.9)* (-5.8)* (-0.0) (-0.0) -o-r

Urban -21298 0.634 5641 0.83 -12402 0.625 3032 -8977 0.017 0.84 _r"o

(-5.0)* (32.1)* (1.0) (-2.2)** (5.2)* (0.5) (-22)** (0.1)ZmOm

Notes: See Table 4 for definitions of symbols. Numbers in parentheses are t-values, R denotes adjusted coefficient of <mdetermination, ro

"o*Sifnificant at the one percent level.**Significant at the five percent level, mZ


Table 7ESTIMATED SAVING EQUATIONS, PERMANENT INCOME MODEL:

ALL HOUSEHOLDS BY REGION

Region Const, YP YT R2

I -2625 0.290 0.494 0,73(4.8)* (19.7)* (54,0)*

I I -39 0.218 0.388 0,50(-0.1) (10.5)* (27.4)*

III -7336 0.346 0.546 0,68(-10.0)* (21.4)* (56,0)*

IV -3692 0.281 0.527 0.70

(-7.7)* (21.6)* (70.4)*

V -2560 0.242 0.472 0.58(-5.2)* (12.5)* (36.2)*

V I -3230 0.291 0,622 0,81(-5.7)* (17.7)* (78.6)*

VII -2475 0,339 0.511 0.74

(-6.7)* (25.5)* (56.5)*

V III -2846 0.303 0.479 0.62

(-6.4)* (15.1 )* (34.6)*

IX -7106 0.512 0,737 0.88(-11.5)* (22,6)* (71.5)*

X -8960 0.548 0.803 0.96

(-14.7)* (33.1)* (138,5)*

X l -7506 0.462 0.722 .9.98(12.9)* (26.1)* (85.6)*

X I I - 1804 0.211 0.652 0.82(-2.7)* (9.4)'* (57.8)*

NCR -3865 0.263 0.401 0.75

(-3.6)* (22.8)* (81.6)*

Notes: See Table 4 for definitions of symbols. Numbers in parentheses aret-values. R2 denotes adjusted coefficient of determination.

*Significant at the one percent level.

Table 8 ,,%ESTIMATED SAVING EQUATIONS, PERMANENT INCOME MODEL: o

RURAL AND URBAN HOUSEHOLDS BY REGION

Region/Household Const. YP YT R2 Const. YP YT D D*YP D_YT R2

Region I:Rurat -5693 0.414 0.618 0.78 -1854 0.218 0.361 -4380 0.212• 0.278 0.7

(-9.2)*• (21.0)* (50.0)* (-1.6) (4.2)* (8.8}* (-3.0)* (3.7)* (6.4)*

Urban -887 0.235 0.409 0.73 -2170 0.206 0.243 1567 0.026 0.167 0.73

(-0.8) (11-1)'* (30.7)* (-0.6) (0.9) (0.9) (0.4) (0.1) (0.6)

Region I1:Rurat -2213 0.310 0.386 0.56 -3282 0.360 0.468 t210 -0.059 -:0.143 0.57

(-3.4)* (13.0}* (24.6)* (-3.5)* (9.3)* (20.0)* (0.9) (-1.2) (-4.6)*C

Urban -250 0.214 0.389 0.46 -6949 0.413 0.435, 8690 -0.230 --0.047 0.46 :0Z

(-0.1) (5.1)* (13.0)* (-1.4) (2.4)** (2.9)* (1.5) (-1.3) (-0.3) >F-

Region II1: o-FI

Rural -2635 0.202 0.566 0.58 -2312 0.184 0.428 -210 0.012 0.176 0.59(-3.0)* (7.0)* (33.3)* (-I .8) (3.6)* (11.6)* (-0.1) (0.2) (4.2)* F

Urban -11374 0.414 0.556 0.70 -9070 0.328 0.320 --2165 0.083 0.240 0.71(--8;9)* (18.6)* (41.6)* (-1.6) (2.2)** (2.9)* (-0.4) "(0.5) (2.2)** z, FFI

Region IV: o• m

RuraF -t235 0.215 0.416 0.52 -2459 0.279 0.461 1808 -0.080 -0.055 0.52 <• I"11

(-2.4)**(11.0)* (11.0)* (39.2)* (-2.5)* (5.6)* (16.7)* (1.4) (-1.4) (-1.8) f-o"0

Urban' -6372 0.330 0.567 0.75 -5028 0.279 0.332 -t478 0.052 0.239 0.76R'I

(-7.7)* 17.6)* (56.7)* (-2.0)** (2.7)* (3.7}* (-0.6) (0.5) (2.7)* z--I

0_Table 8 (continued) _>

c

Region/Household Const. YP YT /72 Const. YP YT D D*YP D*YT R2_>

Region V:Rural -2712 0,247 0.473 0.57 -531 0.086 0.276 -2704 0.188 0.246 0.59

(--5.5)* (9.3)* (30.0)* (-0.7) (1.7) (8.3)* (--2.5)* (3.0)* (6.5)* _om

Urban -2423 0.240 0.483 0.58 3764 -0.154 0.139 -6306 0.405 0.355 0.59 :___(-1.7) (6.2)* (19.8)* (1.0) (-1.0) (1.1) (-1.6) (2.5)* (2.8)* o_

3>Region VI: <

Rural -3423 0.313 0.481 0.66 -3512 0.322 0.421 -84 -0.012 0.108 0.66 c_(--81.)* (16.8)* (39.4)* (-5.2)* (8.6)* (23.0)* (-0.1) (-0.3) (4.4)* =rfl

Urban -5232 0.330 0.651 0.84 -5265 0.422 0.491 -1241 -0.078 0.174 0.84 >'r<

(--3.5)* (11.2)* (51.3)* (-1.8) (5.1)* (12.0)* (--0.4) (-0.9) (4.0)*"rl

Region VI1:Rurat -1732 0.311 0.327 0.63 -500 0.167 0.262 -912 0.153 0.062 0.64

(--6.0)* (20.5)* (30.1)* (-1.0) (4.1)* (8.1)* (-1.4) (3.4)* (1.8)

Urban -4143 0.378 0.572 0.79 - 1275 0.269 0.308 -3401 0.117 0.266 0.79(-4.9)* (16.9)* (41.0)* (-0.4) (1.5) (1_) (-1.1) (0.6) (I .5)

RegionVIII:Rural -2429 0.274 0.457 0.58 -1695 0.199 0.373 150 0.052 0.106 0.59

(-5.7)* (11.3)* (27.6)* (-2.5)* (4.1)* (tl.0)* (0.1) (0.9) (2.7)*

Urban -2974 0.312 0.507 0.63 -2825 0.243 0.258 262 0.061 0.256 0.63(-2.4)** (7.9)* (18.9)* (-1.0) (1.6) (1.6) (0.1) (0.4) (1.6)

RegionIX:Rural --8410 0.604 0.761 0.92 -5085 0.443 0.514 -9780 0.291 0.270 0.93 ._

(-14.5)* (27.6)* (75.9)* (-5.0)* (7.5)* (i5.8)* (-6.3)* (4.4)* (7.9)* =

Tabte8 (continued)

Rqion/Househotd Const. YP YT R2 Conn. YP YT D D*YP D*YT R2

Urban -4414 0.340 0.507 0.55 -2212 0.198 0.362 -2356 0.151 0,156 0.54

(-2,6)* (6.9)* (13.6)* (-0.7) (13) (2.8)* (-0.6) (0.9) (1.1)

Region X:Rural -3811 0.338 0.459 0.66 -2699 0.289 0.449 -3077 0.113 0.029 0.56

"(-7.8)* (16.9)* (29.6)* (-4.2)* (9.7)* (20.2)* (-2.9)* (2,7)* (0.9)

Urban -12655 0.616 0.821 0,97 -162 0,106 0,186 -14424 0.529 0.636 0.97

(-9.1)* (25.9)* (96.2)* (-0.1) (0.6) (1.2) (-3.8)* (3.2)* (4.2)*

Region XI:Rural -696 0.188 0.369 0.53 -404 0.180 0.410 -1781 0_55 --0.089 .5

(-1.2) (7.9)* (25.5)* (-0.5) (5.1)* (20.5)* (-1 A) (1.1) (-3.1)*c

Urban -11964 0.549 0.749 0.90 -8748 0.486 0.499 -3778 0£)64 0.256 0.90Z

(--10.0)* (22,1)* (63.1)* (--2.7)* (4.9)* (5.7)* (--1.1) (0,6) (2.9)* >r-

Region XII: o"11

Rural -202 0.127 0.417 0.48 -1782 0.221 0,464 1735 -0.106 -0.080 0.48 -_I

(0.3) (4.3)* (21.8)* (-1.4) (3.2)* (15.7)* (1.0) (-1.3) (-2.1)** F

Urban -8792 0.406 0.699 0.86 -9366 0.595 0.656 -549 -0.186 0.046 0.86

(-5.2)* (10.7)* (34.3)* (-2.3)** (4.3)* (5.0)* (-0.1) (-1.3) (0.4) zmo

-- m

Notes: See Table 4 for definitions of symbols. Numbers in parentheses are t-values. R2 denotes adjusted coefficient of <mdetermination. _3

73

* Significant at the one percent level.111

** Significant at the five percent level.


components of current income. The importance of lifetime resources,rather than just current income, in the determination of householdsavings (both rural and urban) in the Philippines is thereforeindicated.

From Table 7 the regional MPS for "all" householdsrangesfrom 0.218 to 0.548 out of YP and from 0.388 to 0.803 out of YT- estimatesthat differ substantiallyfrom the valuesof zero and one,respectively, as postulated in the strict version of the permanentincome hypothesis,The latter result also emergeswhen a distinctionis made between rural and urbanhouseholds(Table 8).

It is evident from Table 8 that there are some significantdifferences from the MPS out of YP and YT estimated for rural andurban householdsin each region. Again, it is not possibleto general:ize on the relative magnitudesof the savingpropensitiesbetween thetwo household classes.We note that rural householdsshow higherMPS out of permanent income in four regions(I, Ii, V and IX) andout of transitory income in three regions (I, III and IX). It alsoappears from the significant coefficient estimates for D*YP andD*YT in severalequations that, given the householdslocation (ruralor urban), the marginalsavingrates differ by main sourceof incomesuch that many of them havea positivesign, againindicating higherMPS valuesfor nonfarm vis-a-visfarm households.

A final set of regressionsdistinguishesthree incomegroups(first,secondand third terciles) among rural and urbanhouseholdsin eachregion. Table 9 summarizesthe resultingMPSestimate from the per-manent income model.1O A striking observation is the markedlyhigher MPS out of either permanent or transitory income for thehigher income group. This is especiallytrue among rural households,which show comparable average MPS estimates for the low- andmiddle-income groups that are only about one-half for the high-income group. In the case of urban households,the MPS firstincreasessharply from the smallestvalue for the low-income grouptoan intermediate value for the middle-income group before climbingto the largestvalue for the high-incomegroup.

A possible explanation for this comparative savingbehavior ofrural and urban householdswould be as follows. As indicated above,there is a wide differential between the averageincomesof rural andurban households in the Philippines. Because middle-income ruralhouseholdsalso havegenerally low income levels(compared not onlyto the high-income rural household group but alsoto middle-incomeurban households), they presumably face conditions inimical to

10. For a full presentation of the regression results (involving 144 estimated savingequations in all), see Lamberte and Bautista (1989).


Table9REGIOALAVERAGEVALUESOF ESTIMATEDMAPSOUTOFPERMANENTANDTRANSITORYINCOMEBY RURALAND

URBANHOUSEHOLDSAND BYINCOMEGROUP

Low income Middle income High income All incomegroups

MPS out of PR

Rural 0.16 0.15 0.34 0.22Urban 0.11 0.27 0.38 0.25

MPSout of YT

Rural 0.24 0.27 0.53 0.35Urban 0.14 0.30 0.56 0.33

Note: Eachentry representsthe simpleaverageof MPSestimatesfor the 12regionsbasedontheregressionof Son YPandYT.

savingto a similar extent as do low-income households(e. g., relatingto investment opportunities, coping with borrowing costs,and con-sumption needs).

The averageMPS for low-income rural householdsout of eitherpermanent or transitory income is seen in Table 9 to be higher re-lative to their urban counterparts. While the averageMPS estimatesare higher for urban householdsin the other income categories,theyare not significantly different (based on the two-tailed t-test) fromthose for rural householdsexcept in the middle-income group, andthen only out of permanent income (0.27 versus 0.15). Likewise, theMPS estimates for rural and urban householdsaveraged acrossallregions and income groups are not significantly different. It seemsreasonableto infer from all this, consideringthe much lower averageincome of rural households relative to urban householdsin each _ofthe three income categories,that in general rural householdsin thePhilippines have not only a higher average savin& rate (as shownearlier) but also a higher marginal savingrate than urban householdsat the same income level.


V. CONCLUSIONS

The empirical findings of this study can bebriefly summarized,and some inferencescan be made from them, asfollows:

(1) "Income" isthe most important determinant of householdsavingin the Philippines.This result is true for all the alter-native measuresof current income and its permanent andtransitory components, to the inclusion or exclusion ofother explanatory variables,and for all differences in theclassificationof householdsused(by region, by rural-urbanand by income class).

(2) Lifetime factors, as representedin the permanent incomemeasureby certain household characteristics,have a signi-ficant influence on household savings.The findings alsobear out the hypothesisthat the marginal propensity tosaveout of transitory income is higher than out of perma-nent income.

(3) Marginal saving rates vary widely among households indifferent regions between rural and urban householdsandamong different income classes.The aggregatesavingrateis therefore subject to change as income is redistributedacrossdifferent householdclasses.

(4) The marginal propensity to save of households in theMetro Manila area is estimated to be lower than in any ofthe country's other twelve regionsexcept one. This wouldseem to suggest that a reversal of the past pattern ofregional income growth biasedtoward Metro Manila canpossitivelyaffect the aggregatesavingrate.

(5) It is difficult to generalize about the relative size of themarginal savingrates between rural and urban households.This contracts with the invariably higher average savingrates observedfor urbanhouseholdsin the various regions,attributable to their higher incomesrelative to rural house-holds. However, the estimated marginal saving rates forrural households in many regions are higher than theirurban counterparts, indicating a substantial scope forincreasedsavingswith risingrural income in the context ofagriculture baseddevelopment.

(6) By main source of income, nonfarm households in therural areas tend to save'more, in the "margin,than farm


(agricultural) households. An increasing share of non-agriculture in rural income over time, as the intersectorallinkage effects of agricultural growth work themselvesout, can then lead to a higher marginal saving rate amongrural households.

(7) At a given income level, rural households generally savemore, or dissave less, than urban households, both onaverage and at the margin. Moreover, the marginal savingrate of rural households increases more rapidly as theymove up from the low- and middle-income groups to thehigh-income group compared to their urban counterparts.Under these conditions, even if the average income andsavings of urban households were initially higher, fasterincome growth among rural households will not necessarilyresult in a lower aggregatesaving rate.

In sum, the empirical evidence presented in this paper doesnotprovide support to the notion that the adoption of an agriculture-based development strategy and the associatedshift in rural-urbanincome distribution inevitably entail some sacrifice in domesticsavingsand, thereby, in capital formation and the sustainability ofeconomic growth. Indeed, the observed large potential for expanded

_- rural savings is likely to be realized as the income prospects and in-vestment opportunities are improved for rural households by in-creased public investment in the rural areas and by reduced agri-cultural price distortions. Nothing definite can be predicted, ofcourse, until one does a detailed general equilibrium analysis, takinginto systematic account the various factors affecting aggregatesavings.

As a final point, the macroeconomic benefits of savings in adeveloping country context as described above do not enter in thecalculation of individuals or households acting in isolation, so thataggregate private savings are likely to be lower than is socially desi-rable (Sen 1967). To deal with this externality, "what is needed isnot additional public sector saving, but a subsidy to saving, presu-mably in the form of a higher marginal return" (Deaton 1988: 41).In fact there is an antisavings bias in many LDC government marketinterventions that repress financial intermediation, keep interestrates low, and reduce investment opportunities (McKinnon 1973).Such policy-induced sources of undersaving in the Philippines, asdiscussed in Tan (1981)and Lamberte and Lira (1987), need to beaddressed first before any ambitious government program of sa_ingsmobilization can be rationalized.


Annex INAMES OF REGIONS AND PROVINCES

AND CHARTERED CITIES IN EACH REGION

Region Name Provinces/chartered cities

I Ilocos Abra, Benguet, Ilocos Norte, Ilocos Sur, LaUnion, Mountain Province, Pangasinan

II Cagayan Valley Batanes, Cagayan, Ifugao, Isabela, Kalinga-Apayao, Nueva Vizcaya, Quirino

III Central Luzon Bataan, Bulacan, Nueva Ecija, Pampanga,Tarlac, Zambales, Angeles City, Olongapo

City

IV Southern Tagalog Batangas, Cavite, Laguna, Marinduque,Occidental Mindoro, Oriental Mindoro,Palawan, Quezon, Rizal, Romblon, Aurora

V Bicol Albay, Camarines Norte, Camarines Sur,Catanduanes, Masbate, Sorsogon

VI Western Visayas Aklan, Antique, Capiz, Iloilo, NegrosOccidental, Iloilo City, Bacolod City

VII Central Visayas Bohol, Cebu, Negros Oriental, Siquijor,Cebu City

VIII Eastern Visayas Eastern Samar, Northern Samar, WesternSamar, Leyte, Southern Leyte

IX Western Mindanao Basilan, Sulu, Tawitawi, Zamboanga delNorte, Zamboanga del Sur, ZamboangaCity

X Northern Mindanao Agusan del Norte, Agusan del $ur, Bukid-non, Camiguin, Misamis Occidental, MisamisOriental, Surigao del Norte, Butuan City,Cagayan de Oro City

XI Southern Mindanao Davao del Norte, Davao del Sur, Davao

Oriental, South Cotabato, Surigao del Sur,Davao City

XII Central Mindanao Lanao del Norte, Lanao del Sur, Maguin-danao, North Cotabato, Sultan Kudarat,

Iligan City

NCR National Capital Manila, Pasig, Quezon City, Caloocan City,Region PasayCity, Makati, Other Metro

178 JOURNAL OF PHILIPPINE•DEVELOPMENT

Annex II

CLASSIFICATION OF RURAL AND URBAN HOUSEHOLDS

There is always an element of arbitrariness in differentiating between ruraland urban areas which determines the classification of households into rural andurban. In the 1985 FIES, urban areasare defined to consist of -

1. In their entirety, all cities and municipalities having a population den-sity of at least 1,000 persons per square kilometer.

2. Poblaciones or central districts of municipalities and cities which have apopulation density of at least 500 persons per square kilometer.

3. Poblaciones or central districts (not included in 1 and 2), regardlessofthe population size, which have the following•:

i. street pattern, i.e., network of streets• in either parallel or rightangle orientation;

ii. at least six establishments (commercial, manufacturing, recreationaland/or personal services);

iii. at least three of the following:

(1) a town hall, church or chapel with religious services at leastonce a month;

(2) a public plaza, park, or cemetery;

• (3) a public market or building where trading activities are carriedon at least once a week;

(4) a public building like a school, hospital, puericulture andhealth center or library.

4. Barangays •havingat least 1,000 inhabitants which meet the conditionsset forth in 3 above, and where the occupation of the inhabitants is

predominantly nonfarming or nonfishing.

All areas not falling under any of the above classifications are consideredrural.


Annex IIIESTIMATING PERMANENT INCOME

We follow Bhalla (1980) and Hyun et aL (1979), among others, Jn using anearnings function on which to basethe estimation of permanent income for eachhousehold. The procedure basically involves regressing disposable income onvarious indicators of earning capacity of the household, and the predicted valueis taken as the measure of permanent income. The earning capacity of a house-hold is assumed to be related to its stock of human capital, and physical andfinancial assets.

The available human capital of a household may be gauged in terms of itseducational background, occupational status, and the proportion of members

employed. We therefore included the following explanatory variables in theregression equation for household earning capacity: educational attainment ofhousehold head, represented by the five education categories distinguished inthe 1985 FIES; main source of income, classified into fifteen occupational

categories; and the dependency ratio, which is expected to have a negativeeffect.

FIES data do not include ownership of physical and financial assets.Whatwe have done is represent these stock variables by proxies for flow variables thatcan be identified with the asset values. Available data on purchasesof consumerdurables are used to represent the ownership of physical assets. As to financialassets, the following income flow data are used as proxies: inheritance receivedduring the year, pension and retirement benefits, workmen's compensation,social security benefits, dividends from investments, and profits from the sale of

stocks. The assumption is that the higher the income from such sources, thelarger isthe value of financial assetsheld by the household.

The estimated earnings equations for the thirteen regions are deemed gene-rally satisfactory based on the expected signs of the coefficients, t-values, andadjusted coefficients of determination. Copies of the regression results are avail-able upon request from the authors.


REFERENCES

Adelman, I. "Beyond Export-Led Growth." World Development 12 No. 9(1984).

Alamgir, M. "Rural Savings and Investment in Developing Countries: SomeConceptual and Empirical Issues." Bangladesh Development Studies4, No. 1 (1976).

Bautista, R. M. "Agriculture-Based Development Strategy: Issues and PolicyFramework:" Paper presented at a seminar of the Development PolicyGroup, Resource Systems Institute, East-West Center, Honolulu, onJuly 20, 1988.

B_ry, A. "On Trends in the Gap Between Rich and Poor in Less DevelopedCountries: Why We Know So Little." Review of Income and Wealth31, No. 4 (1985).

Bhalla, S.S. "The Measurement of Permanent Income and its Application toSavings Behavior." Journal of Political Economy 88, No. 4 (1980).

Burkner, H. P. Savings Behavior and Savings Mobilization In Developing Coun-tries. Bochum: Studienverlag Brockmeyer, 1981.

Chowdhury, N. "Household Savings Behavior in Bangladesh: Issues and Evi-dence." Bangladesh Development Studies 15, No. 3 (1987).

Deaton, A. "Saving in Developing Countries." Paper presented at the FirstAnnual World Bank Conference on Development Economics, Washing-ton, D.C., on April 27-28, 1988.

Freedman, D. S. "The Role of the Consumption of Modern Durables in Eco-

nomic Development." Economic Development and Cultural Change19, No. 1 (1970).

Friedman, M. A Theory of the Consumption Function. Princeton: PrincetonUniversity Press,1957.

Gerovitz, M. "Saving and Development." Handbooh of Development Econo-mics. Edited by H. Chenery and N.T, Srinivasan. Amsterdam: Elsevier,1988.

Gupta, K. L. "On Some Determinants Of Rural and Urban Saving Behavior."Economic Record 46 (1970).

Hyun, K. N., D. W. Adams and L. J. Hushak. "Rural Household Savings Beha.vior in South Korea." American Journal of Agricultural Economics 61(1979).

Lamberte, M. B. and R. M. Bautista. "Comparative Saving Behavior of Ruraland Urban Households: The Philippines, 1985." Final report to theTraining and Development Issues (TDI) Project, National Economic andDevelopment Authority, Manila, 1989.


Lamberte,M. B.and J. Lim." RuralFinancialMarkets:A Reviewofthe Literature."PIDS StaffPaper Series No. 87-02. Makati:PhilippineInstituteforDevelopmentStudies.1987.

Left, N. "Dependency Rates and Saving Rates." American Economic Review59, No. 5 (1969).

Lipton, M. Why Poor People Stay Poor: Urban Bias in World Development.Cambridge: Harvard University Press,1977.

Lluch, C., A. A. Powell and R. A. Williams. Patterns in Household DemandandSaving. Washington, D.C.: World Bank, 1977.

McKinnon, R. I. Money and Capita/In Economic Development. Washington,D.C.: Brookings Institution, 1973.

Mellor, J. W. The New Economics of Growth: A Strategy for India and theDeveloping World. Ithaca: Cornell University Press,1976.

• Moore, B. J. "Domestic Savings in Selected Developing Asian Countries."Economic Staff Paper No. 2. Manila: Asian Development Bank, 1987.

Oshima, H. "Personal Savings in the 1985 Philippine Family Income and Expen-diture Survey." Quezon City: University of the Philippines' School ofEconomics, 1988.

Rodriguez, J. A. and R. L. Meyer. "The Analysis of Saving Behavior: The Caseof Rural Households in the Philippines." PIDS Working Paper Series No.88-20. Makati: Philippine institute for Development Studies, 1988.

Sen, A. "Isolation, Assurance, and the Social Rate of Discount." Quarterly

Journal of Economics 81, No. 1 (1967).

Snyder, D.W. "Econometric Studies of Household Saving Behavior in Developing

Countries: A Survey." Journal of Development Studies 10, No. 2 (1974).

Tan, E. "The Structure and Growth of the Philippine Financial Market and theBehavior of its Major Components." PID$ Worhing Paper Series No. 81-06.Makati: Philippine Institute for Development Studies, 1981.

comparative saving behavior of rural and urban … · the intertemporal nature of the saving...

Documents